Re: [WIP] Fix HAVE_CONFIG_H -Wundef warnings.

["Carlos O'Donell" <carlos at redhat dot com>]

On 05/02/2014 04:28 AM, Will Newton wrote:
> On 2 May 2014 09:24, Carlos O'Donell <carlos@redhat.com> wrote:
> 
> Hi Carlos,
> 
>> Is fixing the HAVE_CONFIG_H -Wundef warnings as easy as this?
>>
>> diff --git a/Makeconfig b/Makeconfig
>> index f965398..64b64fc 100644
>> --- a/Makeconfig
>> +++ b/Makeconfig
>> @@ -1081,6 +1081,9 @@ endif
>>  sysd-rules-targets := $(sort $(foreach p,$(sysd-rules-patterns),\
>>                                          $(firstword $(subst :, ,$p))))
>>
>> +# We always configure glibc such that config.h is available.
>> +defines += -DHAVE_CONFIG_H=1
>> +
>>  # A sysdeps Makeconfig fragment may set libc-reentrant to yes.
>>  ifeq (yes,$(libc-reentrant))
>>  defines += -D_LIBC_REENTRANT
>> ---
>>
>> Followed by the mechanical change of all the `#idfef HAVE_CONFIG_H`
>> to `#if HAVE_CONFIG_H`.
>>
>> Then verify the built binaries are identical and pass the testsuite?
>>
>> We always create a config.h for glibc so this looks correct.
>>
>> I don't want to waste my time fixing all of this if I've got it wrong.
> 
> A reasonable number of these are from files shared with gnulib. The
> gnulib versions of the files have in some cases switched to use
> #ifndef _LIBC instead of #if HAVE_CONFIG_H  but I haven't delved into
> why that is yet.
 
I can ignore the gnulib files, thanks to your nice list, and I can
change only glibc ones to use `#if HAVE_CONFIG_H`, but still define
`-DHAVE_CONFIG_H=1`, that way we're ready for the gnulib update.

I've been looking at the gettext update today, but it's more work
than I have time for right now (repeating myself today).

Cheers,
Carlos.